I am working in semiconductor devices and wondering about the noises that can affect my measurement.
Good points to start are here:
http://www.scholarpedia.org/article/1/f_noise
http://www.cs.cmu.edu/~sensing-sensors/readings/1-over-f_noise_pedagogical_review-0204033.pdf
http://www.materialstoday.com/carbon/news/the-mystery-of-the-1f-noise/
http://mwrf.com/semiconductors/revealed-source-low-frequency-1f-noise
It is caused essentially by recombination effects at defects in the semiconductor volume, the borders of diffusion areas or the material surface.
An empirical description after Hooge is a spectrum with:
C1/f=KF. IAF/fB
where Ntot means the total number of moving charges in the device. The Hooge-Parameter α is a material characteristic.
There is no single or ultimate answer for the origin of 1/f noise. It is just present everywhere and all the time.
For semiconductors, scattering in charge motion or trapping in states with distributed time constants are argued to result in nearly 1/f noise.
The magnitude of the 1/f noise in semiconductor devices is function of the DC bias. One usually uses Hooge equation and estimates the Hooge parameter in it, if the number of carriers is known, or estimates charge trap density in unit 1/(m^3*eV), if trapping is assumed. The Hooge parameter normally is denoted with alpha_H and the trap density with Nt. Other parameters are also used, such as power spectrum density PSD(f) at frequency f=1Hz and noise power = integral(PSD*df) in one frequency decade (the latter is used in standards for noise in resistors). Note that PSD is not really power in [W/Hz], but either current or voltage in square, in unit A^2/Hz or V^2/Hz, respectively. The mathematical bottle neck with 1/f noise is that setting PSD=1/f, then integral(PSD*df)=log(f) diverges at f=0Hz and f=infinite. However, PSDs at f=0Hz and f=infinite are impossible for measurement, because these require infinite time of observation or infinite frequency of the spectrometer (neither meaningful for humans). Thus, the interpretation of the 1/f noise in a limited bandwidth (and time) is what takes place in practice.
The bottom line is the top line. There is no single or ultimate answer for the origin of 1/f noise. It just exists. The noise that now we call 1/f noise, was first measured by Johnson in 1925 and then named “flicker noise” in 1926 by Schottky. The 1/f noise is one type of low-frequency noise (LFN). The LFN in semiconductor and other electrical devices is investigated intensively last 50 years.
Thus, search the literature, as Michal Radziwon advised in the previous answer. The LFN is very interesting subject. Although the 1/f electronic noise has a “birthday” (the aforementioned paper of Schottky), the 1/f fluctuation was present all the times. The earliest that I know is the famous “panta rei” (“everything flows”), attributed to an antic Greek philosopher with the amendment “You cannot step twice in the same river” by another philosopher during the Roman times.
Are you working with MOSFETs or other semiconductor devices?
Origing of 1/f noise is very "esoteric" question, so far there some hypothesis and empirical observations e.g. it seems to "like" surfaces, contacts.
Article Electron-Phonon Coupling as the Source of 1/f Noise in Carbon Soot
Sci Rep. 2019 Jan 30;9(1):947. doi: 10.1038/s41598-018-36544-4.
Electron-Phonon Coupling as the Source of 1/f Noise in Carbon Soot.
Mihaila M1, Ursutiu D2, Sandu I3.
Author information
Abstract
Two 1/f noise peaks were found in a carbon soot resistor at voltages characteristic of Kohn anomalies in graphite. The ratio of the electron-phonon coupling matrix elements at the anomalies calculated from the noise peak intensities is the same as the one obtained from the Raman frequencies. This demonstrates that the electron-phonon coupling is the microscopic source of 1/f noise in carbon soot. A new, very general formula was deduced for the frequency exponent, wherein nonlinearity and dispersion are the only ingredients. The interplay between nonlinearity and dispersion in this formula describes the sublinear-supralinear transitions experimentally observed at both anomalies in the voltage dependence of the frequency exponent. A quadratic dependence of the 1/f noise parameter on the matrix element is proposed and applied to explain the M-shape of the 1/f noise in graphene. We found that the frequency exponent mimics the dependence of the noise intensity in the whole voltage range, while both are the image of the graphite phonon spectrum. This implies that the source of nonlinearity is in the electron-phonon coupling which modulates the slope of the spectrum. It requires the presence of 1/f noise in the thermal noise background of the resistor till phonon frequencies.
- Badges
- Science topic